Vineyards - Quantalab

Transcription

Remote Sensing of Environment 99 (2005) 271 – 287
www.elsevier.com/locate/rse
Assessing vineyard condition with hyperspectral indices: Leaf and canopy
reflectance simulation in a row-structured discontinuous canopy
P.J. Zarco-Tejada a,*, A. Berjón b, R. López-Lozano c, J. R. Miller d, P. Martı́n e, V. Cachorro b,
M.R. González e, A. de Frutos b
a
Instituto de Agricultura Sostenible (IAS), Consejo Superior de Investigaciones Cientı́ficas (CSIC), Alameda del Obispo, s/n 14004-Córdoba, Spain
b
GOA-UVA, Universidad de Valladolid, Spain
c
Centro de Investigación y Tecnologı́a Agroalimentaria de Aragón (CITA), Spain
d
Department Earth and Space Science and Engineering, York University, Toronto, Canada
e
Departamento de Producción Vegetal y Recursos Forestales, Universidad de Valladolid, Spain
Received 11 June 2005; received in revised form 5 September 2005; accepted 7 September 2005
Abstract
Methods for chlorosis detection and physiological condition monitoring in Vitis vinifera L. through accurate chlorophyll a and b content (C ab)
estimation at leaf and canopy levels are presented in this manuscript. A total of 24 vineyards were identified for field and airborne data collection
with the Compact Airborne Spectrographic Imager (CASI), the Reflective Optics System Imaging Spectrometer (ROSIS) and the Digital Airborne
Imaging Spectrometer (DAIS-7915) hyperspectral sensors in 2002 and 2003 in northern Spain, comprising 103 study areas of 10 10 m in size,
with a total of 1467 leaves collected for determination of pigment concentration. A subsample of 605 leaves was used for measuring the optical
properties of reflectance and transmittance with a Li-Cor 1800-12 Integrating Sphere coupled by a 200 Am diameter single mode fiber to an Ocean
Optics model USB2000 spectrometer. Several narrow-band vegetation indices were calculated from leaf reflectance spectra, and the PROSPECT
leaf optical model was used for inversion using the extensive database of leaf optical properties. Results showed that the best indicators for
chlorophyll content estimation in V. vinifera L. leaves were narrow-band hyperspectral indices calculated in the 700 – 750 nm spectral region (r 2
ranging between 0.8 and 0.9), with poor performance of traditional indices such as the Normalized Difference Vegetation Index (NDVI). Results
for other biochemicals indicated that the Structure Insensitive Pigment Index (SIPI) and the Photochemical Reflectance Index (PRI) were more
sensitive to carotenoids C x+c and chlorophyll – carotenoid ratios C ab / C x+c than to chlorophyll content C ab. Chlorophyll a and b estimation by
inversion of the PROSPECT leaf model on V. vinifera L. spectra was successful, yielding a determination coefficient of r 2 = 0.95, with an
RMSE = 5.3 Ag/cm2. The validity of leaf-level indices for chlorophyll content estimation at the canopy level in V. vinifera L. was studied using the
scaling-up approach that links PROSPECT and rowMCRM canopy reflectance simulation to account for the effects of vineyard structure, vine
dimensions, row orientation and soil and shadow effects on the canopy reflectance. The index calculated as a combination of the Transformed
Chlorophyll Absorption in Reflectance Index (TCARI), and the Optimized Soil-Adjusted Vegetation Index (OSAVI) in the form TCARI/OSAVI
was the most consistent index for estimating C ab on aggregated and pure vine pixels extracted from 1 m CASI and ROSIS hyperspectral imagery.
Predictive relationships were developed with PROSPECT – rowMCRM model between C ab and TCARI/OSAVI as function of LAI, using fieldmeasured vine dimensions and image-extracted soil background, row-orientation and viewing geometry values. Prediction relationships for C ab
content with TCARI/OSAVI were successfully applied to the 103 study sites imaged on 24 fields by ROSIS and CASI airborne sensors, yielding
r 2 = 0.67 and RMSE = 11.5 Ag/cm2.
D 2005 Elsevier Inc. All rights reserved.
Keywords: Hyperspectral remote sensing; Vitis vinifera; Vineyards; Vine; Radiative transfer; Scaling-up; Optical index; RowMCRM
1. Introduction
* Corresponding author. Tel.: +34 957 499 280/+34 676 954 937; fax: +34
957 499 252.
E-mail address: [email protected] (P.J. Zarco-Tejada).
URL: http://www.ias.csic.es/pzarco (P.J. Zarco-Tejada).
0034-4257/$ - see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2005.09.002
Current research efforts in precision viticulture and on the
temporal and spatial monitoring of Vitis vinifera L. show a
growing interest in remote sensing methods due to its potential
272
P.J. Zarco-Tejada et al. / Remote Sensing of Environment 99 (2005) 271 – 287
for estimating vine biophysical variables such as shape, size
and vigor, potential indicators of fruit quality and yield (a full
review of optical remote sensing methods for vineyard
monitoring can be found in Hall et al., 2002).
Successful mapping of vineyard leaf area index (LAI) using
high spatial IKONOS satellite imagery was shown by Johnson
et al. (2003), enabling the monitoring of plant growth for
irrigation support and canopy management through temporal
relationships between Normalized Difference Vegetation Index
(NDVI) and Leaf Area Index (LAI) (Johnson, 2003). This
index and other ratios were tested by Dobrowski et al. (2002;
2003) such as the Perpendicular Vegetation Index (PVI) and
the Ratio Vegetation Index (RVI) derived from field data and
multispectral aerial photography to estimate canopy density
and dormant pruning weight prediction, suggesting consistency
across growing seasons. As a result of these and other studies,
broad-band multispectral remote sensing imagery of high
spatial resolution shows potential applications for vineyard
canopy structure characterization, leading to a successful
estimation of vine canopy size, shape and row identification
(Hall et al., 2003), vineyard mortality detection and missing
vinestock recognition (Lagacherie et al., 2001), vineyard
classification methods (Lanjeri et al., 2001), and vine canopy
cover estimation for water management (Montero et al., 1999).
These studies point toward the application of new techniques in
viticulture based on precision agriculture, introducing methods
focused on describing homogeneous management zones
derived from remotely sensed biophysical variable estimates
(Hall et al., 2002), connecting the within-field variability and
the suggested classification of the field into different vigor
classes with a potential wine quality production (Johnson et al.,
2001).
Nevertheless, and despite the cited work conducted mainly
with aerial photography, analogue camera systems, and digital
sensors with a limited number of broad bands, little progress
has been made on the remote sensing detection of vineyard
physiology and condition due to the specific characteristics of
the sensors needed, requiring simultaneous narrow-band
capabilities and high spatial resolution. Progress on crop
condition in vineyards has been made at the leaf-level studying
absorptance in the visible region in the field (Schultz, 1996),
and detecting phenology and grape color at harvest to gather
information about berry phenolics (Lamb et al., 2004). For this
reason, and due to such limited work achieved on the remote
sensing of vine physiology and condition at canopy level,
current research efforts are warranted toward the investigation
of physical methods applied to high-spatial resolution hyperspectral remote sensing imagery to estimate leaf biochemical
constituents and canopy biophysical variables to gather
information related to vine status and functioning (ZarcoTejada et al., 2003). Several studies indicate that the estimation
of leaf biochemistry may be used as indicators of chlorosis due
to plant stress and nutritional deficiencies caused by micro and
macro elements (Fernandez-Escobar et al., 1999; Jolley &
Brown, 1994; Marschner et al., 1986; Tagliavini & Rombolà,
2001; Wallace, 1991). As an example, element deficiencies
such as iron and nitrogen may result in vine chlorosis and may
cause a decrease of fruit yield and quality in the current and the
subsequent year as fruit buds develop poorly (Tagliavini &
Rombolà, 2001).
Leaf biochemistries, such as the concentration of chlorophyll a + b (C ab), water (C w), and dry matter (C m), are
indicators of stress and growth that may be estimated by
empirical methods (indices) and analytical techniques (physical
methods) from remote sensing data in the 400– 2500 nm
spectral region. Several studies demonstrate the feasibility of
chlorosis detection in vegetation through C ab estimation using
spectroscopy and leaf optical properties (Carter & Spiering,
2002; Gitelson et al., 2003; Jacquemoud et al., 1996; le Maire
et al., 2004; Sims & Gamon, 2002). Recently, several new
optical indices have been proposed to relate crop physiological
status with hyperspectral data through their relationship to
biochemical constituent concentrations such as chlorophyll
(Carter, 1994; Gitelson & Merzlyak, 1996; Vogelmann et al.,
1993; Zarco-Tejada et al., 2001, 2004, 2005), carotenoids
(Fuentes et al., 2001; Sims & Gamon, 2002), and water content
(Gao, 1996; Peñuelas et al., 1997).
A large number of the new narrow-band optical indices that
might be used with leaf and canopy hyperspectral reflectance
have been tested on specific crop and forest species with
success (Haboudane et al., 2002, 2004; Zarco-Tejada et al.,
2001; a full review of indices can be found in Zarco-Tejada et
al., 2005). Red edge reflectance indices, spectral and derivative
indices, and derivative ratios have demonstrated good results
for C ab estimation from canopy reflectance using airborne
hyperspectral data. Recently, combinations of indices based on
the Transformed Chlorophyll Absorption in Reflectance Index,
TCARI (Haboudane et al., 2002), Modified Chlorophyll
Absorption in Reflectance MCARI (Daughtry et al., 2000),
and the Optimized Soil-Adjusted Vegetation Index, OSAVI
(Rondeaux et al., 1996), such as TCARI/OSAVI and MCARI/
OSAVI, have been demonstrated to successfully minimize soil
background and LAI variation in crops, providing predictive
relationships for precision agriculture applications with hyperspectral imagery in closed crops (Haboudane et al., 2002) and
open tree canopy orchards (Zarco-Tejada et al., 2004).
Nevertheless, and despite the successful relationships obtained
between specific optical indices and leaf biochemistry in closed
crops, estimation of such biochemical components in vineyards
at a canopy level from remote sensing requires appropriate
modeling strategies accounting for its row-structure and large
shadow and soil effects on the bi-directional reflectance
(BRDF) signature. The application of previously validated leaf
optical indices to discontinuous crop canopies such as V.
vinifera L. needs extensive research with airborne hyperspectral data of optimum spatial and spectral resolution, i.e. one
meter or better spatial resolution to obtain pure vine reflectance
in selected spectral bands sensitive to pigment absorption. Vine
row geometry leads to large variations in shadow scene
proportions as a function of sun azimuth and zenith angles
relative to row orientation, affecting the vegetation index and
the estimated leaf biochemical constituent. It is required,
therefore, that successful vine leaf-level indices are investigated at the canopy level through scaling-up simulation using
appropriate physical methods and very high spatial resolution.
This approach (Haboudane et al., 2002; Zarco-Tejada et al.,
2001, 2003) uses physical models at leaf and canopy levels to
scale-up optical indices that are sensitive to specific biochemical constituents, therefore modeling the indices as function of
canopy structure, viewing geometry and background effects.
This manuscript reports on a study of the optical properties
of V. vinifera L. for C ab estimation using narrow-band indices
and radiative transfer model inversion. The work describes in
detail the methods for accurate measurements of the leaf optical
properties, testing the behavior of several indices for successful
C ab estimation. The successful leaf optical indices are proposed
for scaling-up simulation with the Markov-Chain Canopy
Reflectance Model (MCRM) (Kuusk, 1995a,b) with additions
to simulate the row crop structure, called rowMCRM, and
developed within the frame of the Crop Reflectance Operational Models for Agriculture (CROMA) project. The linked
PROSPECT – rowMCRM model is assessed to model vineyard
scene component proportions, row orientations, vineyard
dimensions and background effects with high spatial resolution
hyperspectral airborne imagery.
273
study area at the time of data acquisition to derive aerosol
optical depth at 550 nm. Atmospheric correction was applied to
ROSIS radiance imagery using MODTRAN, whereas the
CAM5S atmospheric correction model (O’Neill et al., 1997)
was used for CASI imagery. Reflectance data were georeferenced using GPS data collected onboard the aircraft. Soil
reflectance spectra were used to perform a flat-field correction
(Ben-Dor & Levin, 2000) that compensated for residual effects
on derived surface reflectance estimations in atmospheric water
and oxygen absorption spectral regions. Fig. 1 shows
vegetation and soil spectra extracted from CASI mapping
mission image on selected sites after processing to surface
reflectance, observing the large variability in soil brightness
levels, as well as the pure vine and the mixed soil + vine +
shadow spectra.
Concurrent with the airborne overflights, field sampling
campaigns were conducted in summer 2002 and 2003 for
biochemical analysis of leaf C ab, as well as to measure
reflectance (R) and transmittance (T) from leaf samples to
study the vine optical properties.
2.2. Study site description and leaf sampling methods
2. Airborne and field campaigns for data collection
2.1. Airborne campaigns with ROSIS and CASI hyperspectral
sensors
Data acquisition campaigns were conducted in July 2002
under the European Union HySens-2002 project intended to
investigate physical methods with the Reflective Optics System
Imaging Spectrometer (ROSIS) and the Digital Airborne
Imaging Spectrometer (DAIS-7915) airborne hyperspectral
sensors to estimate leaf biochemical constituents in vineyard
canopies. In July 2003 the Compact Airborne Spectrographic
Imager (CASI) sensor was flown over Spain in collaborative
research with York University (Canada) and the Spanish
aerospace institute Instituto Nacional de Técnica Aeroespacial
(INTA). Both campaigns took place in study areas of V.
vinifera L. in Ribera del Duero D.O. in Northern Spain.
ROSIS imagery were acquired at 1 m spatial resolution, and
calibrated to at-sensor radiance by the German Aerospace
Center (DLR). CASI imagery were collected on two airborne
missions, each with a specific sensor mode of operation: i) the
Mapping Mission, with 1 m spatial resolution and 8 userselected spectral bands placed in the spectrum to enable the
calculation of specific narrow-band indices sensitive to
pigment concentration (bands were centered at 490, 550,
670, 700, 750, 762, 775 and 800 nm with full-width at half
maximum (FWHM) ranging between 7 and 12 nm); and the
Hyperspectral Mission, with 4 m spatial resolution, 72
channels and 7.5 nm spectral resolution. The 12-bit radiometric
resolution data collected by CASI were processed to at-sensor
radiance using calibration coefficients derived in the laboratory
by the Earth Observations Laboratory (EOL), York University,
Canada. Aerosol optical depth data at 340, 380, 440, 500, 670,
870, and 1020 nm were collected using a Micro-Tops II
sunphotometer (Solar Light Co., Philadelphia, PA, USA) in the
The study sites of V. vinifera L. used for ground and
airborne collection were carefully selected from a plot network
currently monitored by the local government to assure a
gradient in the leaf biochemistry as sought for this study. A
total of 10 fields were selected in 2002 and 14 fields in 2003
for leaf sampling collection, comprising a total of 103 study
areas of 10 10 m in size. In 2002, 10 leaves per site were used
for Cab sampling and reflectance and transmittance measurements. In 2003, a total of 80 leaves were sampled from each
10 10 m study area, using 50 leaves for measuring dry matter
and elements N, P, K, Ca, Mg, Fe, 20 leaves for Cab
determination, and 10 leaves per site for conducting reflectance
and transmittance measurements. A total of 1467 leaves were
used for determination of Cab on the 103 study sites comprised
by the 2002 and 2003 campaigns, and 605 leaves for measuring
the optical properties. Leaves used for measuring optical
properties were taken to the laboratory and reflectance and
transmittance measurements made on the same day to avoid
pigment degradation. Dry matter was measured placing the
samples in a pre-heated oven at 40 -C until a stable dry weight
was reached. Structural measurements on each study site
consisted of grid size, number of vines within the 10 10 m
site, trunk height, vegetation height and width, and row
orientation. Soil samples were collected at each site for
laboratory analysis. Fig. 2 illustrates a CASI image acquired
from vineyard fields, showing 15 out of the total 103 blocks of
10 10 m used for leaf sampling and ground data collection.
2.3. Leaf pigment determination by destructive sampling
The leaves from the V. vinifera L. sites were sampled from
the top of the canopy, eliminating the small leaves indicative of
low expansion. Leaves were placed in paper bags to allow
tissue respiration and conservation, then stored at 4 -C prior to
274
Canopy Reflectance
0.5
0.4
Soil (brighter)
Soil (darker)
Pure Vine
Mixed (soil+veg)
0.3
0.2
0.1
0
400
500
600
700
800
900
Wavelength (nm)
Fig. 1. Airborne CASI image collected at 1 m spatial resolution and 8 bands selected for calculation of narrow-band indices sensitive to pigment concentration. Pure
vine reflectance and soil spectra extracted from the image show the within-field variability.
analysis of reflectance and transmittance, and then stored in a
freezer at 8 -C prior to pigment determination. A 1.6 cm
circle from each leaf sample was cut out for grinding with 4 ml
acetone at 80%, and adding 8 ml acetone to a total of 12 ml in
each tube. Tubes were stored in the dark at 4 -C for 48 h prior
to spectrophotometer measurements. Each sample for pigment
determination was filtered, placed in a cuvette and the
absorbance measured between 400 and 700 nm with 2 nm
fixed resolution at 1 nm interval with a Jasco V-530 UV-VIS
spectrophotometer (Jasco Inc., Great Dunmow, UK). Chloro-
Fig. 2. Airborne hyperspectral CASI image acquired from one of the Vitis vinifera L. fields in this study, showing 15 blocks of 10 10 m selected for leaf sampling
and ground data collection.
Table 1
Range of variation for the leaves sampled from the 103 study sites of Vitis
vinifera L. used in this study
Ca
Cb
C ab
C x+c
C a / C b C a / C x+c C b / C x+c C ab / C x+c
Max
54.03 19.87 73.45 13.98 4.29
Min
1.74 1.25 3.40 1.39 1.05
Average 26.01 8.98 34.99 7.90 2.92
5.08
0.74
3.20
1.77
0.33
1.11
6.75
1.06
4.31
Values in Ag/cm2 (n = 1467).
phyll a (C a), chlorophyll b (C b), and total carotenoid (C x+c )
concentrations were calculated using the extinction coefficients
derived by Wellburn (1994) and the absorbance measured at
470, 646, and 663 nm with Eqs. (1) –(3).
Ca ¼ 12:21IA663 2:81IA646
ð1Þ
Cb ¼ 20:13IA646 5:03IA663
ð2Þ
Cxþc ¼ ð1000IA470 3:27ICa 104ICb Þ=198:
ð3Þ
These measurements resulted in a mean C ab of l = 34.99 Ag/
cm2, with a wide range between 3.4 and 73.45 Ag/cm2
(n = 1467). Table 1 shows the range of variation of C a, C b,
C ab and C x+c used later for determination of pigment content at
each study site. The range of variation of the subset of 605
leaves used for optical measurements, correlation with optical
indices and modelling are shown in Table 2.
2.4. Protocol for optical measurements of V. vinifera L. leaves
Reflectance and transmittance measurements of vine leaves
were conducted on the subsample of 605 leaves with a Li-Cor
1800-12 Integrating Sphere (Li-Cor, Inc., Lincoln, NE, USA),
coupled by a 200 Am diameter single mode fiber to an Ocean
Optics model USB2000 spectrometer (Ocean Optics Inc.,
Dunedin, FL, USA), with a 2048 element detector array, 0.5
nm sampling interval, and 7.3 nm spectral resolution in the
350 – 1000 nm range. Software was designed for signal
verification, adjustment of integration time, and data acquisition. An integration time of 13 ms was used for all sample
measurements. Spectral bandpass characterization performed
using a mercury spectral line lamp source yielded FWHM
bandwidth estimates of 7.3 at 546.1 nm. Fiber spectrometer
wavelength calibration was performed using the Ocean Optics
HG-1 Mercury –Argon Calibration Source that produces Hg
and Ar emission lines between 253 and 922 nm. Single leaf
reflectance and transmittance measurements were acquired
following the methodology described in the manual for the LiCor 1800-12 system (Li-Cor Inc., 1984) modified by Harron
Table 2
Range of variation for the subsample of Vitis vinifera L. leaves used for optical
measurements and correlations with optical indices
Ca
Cb
C ab
C x+c
C a / C b C a / C x+c C b / C x+c C ab / C x+c
Max
53.79 19.49 70.83 13.98 4.29
Min
1.74 1.25 3.40 1.39 1.05
Average 25.43 8.80 34.23 7.99 2.91
Values in Ag/cm2 (n = 605).
4.31
0.96
3.09
1.67
0.35
1.08
5.98
1.50
4.17
275
(2000) to correct for stray-light in the integrating sphere. For
clarity, the protocol is described here with the steps required to
calculate the stray-light corrected leaf hemispherical reflectance
(R) and transmittance (T) using a reference target in the
integrating sphere. The protocol consisted of a total of five
measurements modifying the position of the collimated light,
dark and white plugs in the integrating sphere to measure the
transmittance signal (TSP), the reflectance signal (RSS), the
reflectance internal standard (RTS), the reflectance ambient
(RSA), and the dark measurement (DRK) (Table 3)). For clarity
of the protocol used in this study and for future reference, a
schematic view of the integrating sphere with lamp and port
placement is shown in Fig. 3. Stray-light corrected reflectance
and transmittance were then calculated assuming a constant
center wavelength and spectral bandpass, using the set of
equations described by Harron (2000) for stray-light correction
in broadleaves without the requirement of sample carriers (Eqs.
(4) – (8)). Another measurement protocol and a different set of
equations are proposed in Harron (2000) when measuring
needle samples or broad leaves smaller than the sphere sample
port.
R ¼ RV
RV
GTfi þ RBaSO
Si Rw Tf
ð4Þ
4
GTfi þ Si Rw Tf
Si Rw RTf
T ¼ TV 1 þ
GTfi RBaSO4
ð5Þ
with RV, TVand G given in Eqs. (6), (7) and (8),
RV ¼
RSS RSA
RBaSO4
RTS RSA
ð6Þ
TV ¼
TSP DRK
RBaSO4
RTS RSA
ð7Þ
G ¼ ð1 Wf Rw Bf RBaSO4 Þ
ð8Þ
with the following coefficients measured for the sphere used in
this study,
W f is the fraction of the sphere which is interior wall (0.968)
B f is the fraction of the sphere which is BaSO4 reference
(0.009)
Table 3
Sequence of measurements with the Li-Cor 1800 integrating sphere and fiber
spectrometer to enable the calculation of reflectance and transmittance with
Eqs. (4) – (8) and the schematic view shown in Fig. 3
Step
Setup
Lamp
White plug
Dark plug
Sample
1
2
3
4
5
RSA
RSS
RTS
TSP
DRK
C (ON)
C (ON)
B (ON)
A (ON)
OFF
B
B
C
C
B
A
A
A
B
A
OUT
IN@
IN@
INY
OUT
IN Y: adaxial leaf surface facing sample port A.
IN @: adaxial leaf surface facing sphere.
276
Fiber Optics
attached to
Spectrometer
Dark Plug
B
Sample Port
C
A
Lamp
White Plug
Fig. 3. Schematic view of the Li-Cor 1800-12 integrating sphere attached to an USB2000 fiber spectrometer used for reflectance and transmittance measurements.
Experimental measurements made with the Li-Cor 1800-12
integrating sphere used in this study were employed to calculate
the fraction of light scattered into the sphere from the
illuminator (S i), a function of the optical properties of the
sphere and light lamp used. The S i value obtained for the sphere
used in this study was 10 4, indicating that such correction
factor had a very small effect on the calculated reflectance and
transmittance. Nevertheless, on another integrating sphere used
on this same study with different design and configuration (data
not shown), the S i correction was critical for an accurate
calculation of the leaf optical properties corrected for straylight, and therefore it has to be seriously taken into consideration in such cases. As an example of spectra measured with this
methodology, Fig. 4 shows vine leaf reflectance and transmittance spectra with pigment content of 26.68 Ag/cm2, and 4 leaf
reflectance measurements from leaves containing a gradient in
chlorophyll concentration between 15 and 54 Ag/cm2.
stressed vegetation in the visible and the red edge spectral
region (Carter, 1994; Carter & Spiering, 2002; Horler et al.,
1983; Vogelmann et al., 1993; Zarco-Tejada et al., 2001). A full
review of these chlorophyll indices can be found in ZarcoTejada et al. (2001, 2004, 2005) and summarized in Table 4.
These indices are generally classified into visible and visible /
NIR ratios, red edge indices and spectral and derivative
0.7
0.6
R (Cab=26.68)
T (Cab=26.68)
0.5
R (T)
T fi is the fraction of the sphere which is target and directly
illuminated (0.006)
T f is the fraction of the sphere which is target (0.009)
S i is the fraction of light scattered into the sphere from the
illuminator (0.0001)
R BaSO4 is the reflectance of the BaSO4 reference (0.98)
R w is the reflectance of sphere walls (0.9).
0.4
0.3
0.2
0.1
0
400
500
600
700
800
700
800
Wavelength (nm)
0.7
0.6
0.5
3. Vegetation indices and model inversion for C ab
estimation in V. vinifera L. at the leaf-level
R (Cab=15.6)
R (Cab=26.7)
R (Cab=35.4)
R (Cab=53.9)
R
0.4
Leaf-level spectroscopy enables the calculation of narrowband indices potentially related to specific light absorptions
caused by leaf biochemical constituents, such as chlorophyll a
and b (Carter & Spiering, 2002; Sims & Gamon, 2002; ZarcoTejada et al., 2005), carotenoids/chlorophyll and anthocyanins/
chlorophyll ratios (Fuentes et al., 2001; Gamon & Surfus,
1999; Peñuelas et al., 1995), dry matter (Fourty & Baret, 1997),
and water content (Carter, 1991; Ceccato et al., 2001; Danson
et al., 1992; Gao, 1996; Peñuelas et al., 1997). Several optical
indices are currently used with success for C ab estimation from
leaf optical properties on different crop and forest species,
exploiting the differences in reflectance between healthy and
0.3
0.2
0.1
0
400
500
600
Wavelength (nm)
Fig. 4. Sample leaf reflectance (R) and transmittance (T) measured in a Vitis
vinifera L. leaf with integrating sphere and protocol described in this
manuscript. Destructive C ab determination yielded a chlorophyll content of
26.68 Ag/cm2 (upper plot). Bottom plot shows 4 leaf reflectance measurements
for vine leaves containing 15.6, 26.7, 35.4, and 53.9 Ag/cm2 respectively.
277
Table 4
Hyperspectral optical indices used in this study
Vegetation index
Equation
Normalized Difference Vegetation Index (NDVI)
NDVI ¼ ðRNIR Rred Þ=ðRNIR þ Rred Þ
Rouse et al. (1974)
SR ¼ RNIR =Rred
Jordan (1969); Rouse et al. (1974)
Simple Ratio Index (SR)
Modified Simple Ratio (MSR)
Modified Triangular Vegetation Index (MTVI1)
Modified Triangular Vegetation Index (MTVI2)
Renormalized Difference Vegetation Index (RDVI)
Greenness Index (G)
MSR ¼
Reference
Chen (1996)
RNIR =Rred 1
0:5
ðRNIR =Rred Þ
þ1
MTVI1 ¼ 1:2*½1:2*ðR800 R550 Þ 2:5*ðR670 R550 Þ
Haboudane et al. (2004)
1:5*½1:2*ðR800 R550 Þ 2:5*ðR670 R550 Þ
MTVI2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffi
ð2*R800 þ 1Þ2 6*R800 5* R670 0:5
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
RDVI ¼ ðR800 R670 Þ= ðR800 þ R670 Þ
Rougean and Breon (1995)
–
Improved SAVI with self-adjustment
factor L (MSAVI)
G ¼ R554 =R677
TVI ¼ 0:5*½120*ðR750 R550 Þ 200*ðR670 R550 Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
2*R800 þ 1 ð2*R800 þ 1Þ2 8*ðR800 R670 Þ
MSAVI ¼
2
Optimized Soil-Adjusted Vegetation Index (OSAVI)
OSAVI ¼ ð1 þ 0:16Þ*ðR800 R670 Þ=ðR800 þ R670 þ 0:16Þ
Rondeaux et al. (1996)
MCARI ¼ ½ðR700 R670 Þ 0:2*ðR700 R550 Þ*ðR700 =R670 Þ
Daughtry et al. (2000)
Triangular Veg. Index (TVI)
Modified C ab Absorption in Reflectance
Index (MCARI)
Transformed CARI (TCARI)
Broge and Leblanc (2000)
Qi et al. (1994)
TCARI ¼ 3*½ðR700 R670 Þ 0:2*ðR700 R550 Þ*ðR700 =R670 Þ
Haboudane et al (2002)
Modified Chlorophyll Absorption in Reflectance
Index (MCARI1)
Modified Chlorophyll Absorption in Reflectance
Index (MCARI2)
MCARI1 ¼ 1:2*½2:5*ðR800 R670 Þ 1:3*ðR800 R550 Þ
1:5*½2:5*ðR800 R670 Þ 1:3*ðR800 R550 Þ
MCARI2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffi
ð2*R800 þ 1Þ2 6*R800 5* R670 0:5
Zarco and Miller (ZM)
ZM ¼ R750 =R710
Zarco-Tejada et al. (2001)
Blue/Green and Blue/Red Pigment
ndices (RGI, BGI, BRI)
RGI ¼ R690 =R550
Zarco-Tejada et al. (this study)
BGI1 ¼ R400 =R550
BGI2 ¼ R450 =R550
BRI1 ¼ R400 =R690
BRI2 ¼ R450 =R690
Simple Ratio Pigment Ind. (SRPI)
Normalized Phaeophytinization Index (NPQI)
Photochemical Reflectance Index (PRI)
SRPI ¼ R430 =R680
NPQI ¼ ðR415 R435 Þ=ðR415 þ R435 Þ
Peñuelas et al. (1995)
PRI1 ¼ ðR528 R567 Þ=ðR528 þ R567 Þ
Gamon et al. (1992)
Barnes (1992)
PRI2 ¼ ðR531 R570 Þ=ðR531 þ R570 Þ
Normalized Pigment Chlorophyll Index (NPCI)
Carter Indices (CTR)
PRI3 ¼ ðR570 R539 Þ=ðR570 þ R539 Þ
NPCI ¼ ðR680 R430 Þ=ðR680 þ R430 Þ
CTR1 ¼ R695 =R420 CTR2 ¼ R695 =R760
Carter (1994,1996)
LIC1 ¼ ðR800 R680 Þ=ðR800 þ R680 Þ
LIC2 ¼ R440 =R690
LIC3 ¼ R440 =R740
SIPI ¼ ðR800 R450 Þ=ðR800 þ R650 Þ
Lichtenthaler et al. (1996)
Vogelmann Indices (VOG)
VOG1 ¼ R740 =R720
VOG2 ¼ ðR734 R747 Þ=ðR715 þ R726 Þ
VOG3 ¼ ðR734 R747 Þ=ðR715 þ R720 Þ
Vogelmann et al. (1993);
Gitelson and Merzlyak (GM)
Curvature Index (CUR)
GM1 ¼ R750 =R550
Gitelson and Merzlyak (1997)
Lichtenthaler Indices (LIC)
Structure Insensitive Pigment Index (SIPI)
GM 2 ¼ R750 =R700
CUR ¼ ðR675 IR690 Þ= R2683
analysis indices. Other traditional indices related to vegetation
structure and condition, such as NDVI or the Simple Ratio
(SR), normally show low relationships with leaf biochemical
constituents (Zarco-Tejada et al., 2001) and consistently show
unsuccessful performance in detecting physiological stress
condition.
Despite the demonstrated success of these narrow-band leaf
indices for C ab estimation, spectral reflectance signatures from
agricultural canopies are characterized by large contributions
from the soil background and LAI variation at different growth
stages. In these cases, scaling-up methods through canopy
reflectance models are needed to account for crop structure,
viewing geometry and soil and shadow effects on the
reflectance. The study of indices at both leaf and canopy
levels demonstrates that successful indices developed at the
leaf-level do not necessarily perform well at the canopy level
due to the soil and structural effects mentioned (Zarco-Tejada
et al., 2001, 2004). Therefore, combined indices have been
proposed to minimize background soil effects while maximizing the sensitivity to C ab (Haboudane et al., 2002) and LAI
278
(Haboudane et al., 2004) and to yield prediction relationships
directly applicable to hyperspectral imagery. As an example,
CARI (Chlorophyll Absorption in Reflectance Index) (Kim et
al., 1994) was shown to reduce the variability induced on
photosynthetically active radiation inferences due to nonphotosynthetic materials. MCARI (Modified Chlorophyll Absorption in Reflectance Index) (Daughtry et al., 2000) was a
modification of CARI to minimize the combined effects of the
soil reflectance and the non-photosynthetic materials. SAVI
(Soil-Adjusted Vegetation Index) (Huete, 1988) and OSAVI
(Optimized Soil-Adjusted Vegetation Index) (Rondeaux et al.,
1996) were proposed as soil-line vegetation indices that could
be combined with MCARI to reduce background reflectance
contributions (Daughtry et al., 2000). As a result of the
development of these indices, successful C ab estimation on
corn agricultural canopies at different growing stages was
achieved using the TCARI/OSAVI combined index in forward
leaf-canopy modelling, proving its robustness due to the low
sensitivity to effects caused by LAI variation and background
influence (Haboudane et al., 2002). Other indices, such as the
modified chlorophyll absorption ratio index (MCARI2), are
discussed in depth in Haboudane et al. (2004). All these
mentioned indices that can be found in Table 4 have proven
different degrees of success in crop and forest species for
pigment estimation at the leaf-level. Relationships between all
these single and combined indices calculated from the 605 leaf
R and T measurements in 2002 and 2003 campaigns (Table 4)
and pigment content measurements C a, C b, C ab, C x+c , and
pigment ratios C a / C b, C a / C x+c , C b / C x+c and C ab / C x+c (Table
2) were calculated using linear, exponential, and 3rd order
polynomial functions to allow for both linear and non-linear
relationships between indices and leaf pigment concentrations
in grape leaves. The large database available of leaf optical
measurements as part of this study addresses the lack of
previous studies on investigating appropriate leaf optical
indices in V. vinifera L. crop.
In addition to the generally accepted relationships existing
between leaf optical indices and C ab, model inversion
methods using radiative transfer simulation have been
successfully used to simulate leaf optical properties. Due to
its extensive validation, the PROSPECT model (Jacquemoud
& Baret, 1990), based on the plate model (Allen et al., 1969,
1970), was used in this study to simulate the leaf optical
properties of V. vinifera L. leaves, testing the feasibility of
C ab estimation. Several studies demonstrate successful retrievals of pigment concentration from leaf optical properties
with PROSPECT (Jacquemoud & Baret, 1990; Jacquemoud
et al., 1996; le Maire et al., 2004) although limited simulation
work has been conducted with extensive measurements on V.
vinifera L leaves. The large database of leaf reflectance and
transmittance spectra measured in 2002 and 2003, comprising
a total of 605 measurements, were used for PROSPECT
model inversion. The model inversion was performed by
iterative optimization, varying input parameters N (structural
parameter) from 1 to 2.5, C ab between 5 and 95 Ag/cm2, C m
in the range 0.001 and 0.04 mg/cm2, and C w for 0.001 and
0.04 mg/cm2, obtaining the root mean square error (RMSE)
function n(N) to be minimized using both R and T in the
400– 800 nm range using Eq. (9),
RMSE ¼ nð N ; Cab ; Cm ; Cw Þ
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
iffi
u h
u ~ ðRPROSPECT Rm Þ2 þ ðTPROSPECT Tm Þ2
k
k
t
k
¼
n
ð9Þ
where R m and T m are reflectance R and transmittance T
measured from n leaf samples with the Li-Cor integrating
sphere and fiber spectrometer. Estimated C ab values from
each R and T spectra by inversion were then compared with
leaf destructive measurements of pigment concentration, and
the RMSE for C ab estimation from the entire database was
calculated.
4. Application of leaf-level hyperspectral indices at canopy
level in V. vinifera L. with the rowMCRM model
PROSPECT was linked to the rowMCRM model, which
refers to the Markov-Chain Canopy Reflectance Model
(MCRM) (Kuusk, 1995a,b) with additions to simulate the
row crop structure. The rowMCRM model was developed
within the frame of the Crop Reflectance Operational Models
Table 5
Nominal values and range of parameters used for leaf and canopy simulation
with PROSPECT and rowMCRM for the vine study sites
PROSPECT
Leaf parameters
Nominal values and range
Chlorophyll a + b (C ab)
Dry matter (C m)
Equivalent water thickness (C w)
Structural parameter (N)
5 – 95 Ag/cm2
0.0035 mg/cm2
0.025 mg/cm2
1.62
rowMCRM
Canopy layer and structure parameters
Row Leaf Area Index (LAI)
Leaf Angle Distribution Function (LADF)
Relative leaf size (h s)
Markov parameter (k z)
Leaf transmittance coefficient (t)
Leaf hair index (l h)
Canopy height (C H)
Crown width (C W)
Visible soil strip length (V s)
Diff. between sun azimuth and row
direction (w)
1–5
e = 0.95; h n = 45- (plagiophile)
0.083
1.1
0.9
0.1
1.2 – 1.8 m
0.6 – 1.3 m
1.7 – 2.3 m
11 – 95.2-
Background and viewing geometry
Soil reflectance (q s)
Angstrom turbidity factor (b)
Viewing geometry (h sh v /)
From images (Fig. 7)
0.18
Calculated for each image and site
Canopy structural parameters were used in the rowMCRM model for
simulation of the canopy reflectance by radiative transfer. Leaf structural
parameters, and leaf biochemical parameters were used for leaf-level simulation
of reflectance and transmittance using PROSPECT.
0.40
PECT –rowMCRM used in this study are shown in Table 5: i)
leaf parameters for simulating the leaf optical properties, such
as chlorophyll a + b (C ab), dry matter (C m), water content (C w),
and structural parameter (N); ii) canopy layer and structure
parameters such as the row leaf area index (LAI), leaf angle
distribution function (LADF), relative leaf size (h s), Markov
parameter (k z), leaf transmittance coefficient (t), leaf hair index
(l h), canopy height (C H), crown width (C W), visible soil strip
length (V s), and the angular difference between sun azimuth
and row direction (w); and iii) background and viewing
geometry parameters such as soil reflectance (q s), Angstrom
turbidity factor (b), and the viewing geometry (h s, h v, /).
Without an intention to provide an in-depth sensitivity
analysis for rowMCRM (work conducted as part of CROMA
project), an exploratory analysis was carried out to study the
effects of the different input parameters for PROSPECT –
rowMCRM on the canopy reflectance and selected optical
indices. The soil background and distance between rows
(visible soil strip) input for this row-structured canopy are
shown to have large effects on the canopy reflectance (Fig. 5).
As expected, soil brightness effects are greater as function of
the visible soil strip, suggesting the importance of the vineyard
architecture for successful simulation of the canopy reflectance.
Typical vineyard canopies are planted in grids with a distance
between rows of around 2 m (1.7 to 2.3 m range in the 103
study sites in this study), resulting in reflectance differences of
Canopy Reflectance
Bright soil
Dark soil
0.30
0.20
Vs=3
0.10
Vs=1
0.00
400
500
600
700
800
Wavelength (nm)
Fig. 5. Canopy reflectance simulation conducted with PROSPECT –
rowMCRM as function of soil background (q s) and visible soil strip (V s).
for Agriculture project (CROMA), with a goal to successfully
simulate different scene component proportions, as a function
of row orientations and crop dimensions, and soil background
and shadow effects as function of viewing geometry in rowstructured crop canopies. Therefore the rowMCRM canopy
reflectance model was considered an optimum candidate for
the scaling-up of narrow-band indices in row-structured
vineyard canopies. The inputs required to the link PROS-
0.04
0.04
LAI=1
0.03
TCARI/OSAVI
TCARI/OSAVI
LAI=5
CW=1.5
0.03
CW=1.5
0.02 C =1
W
CW=0.5
0.02
CW=1
CW=0.5
0.01
0.01
Ch=3
Ch=1.5
Ch=0.5
0.00
0
1
Ch=3
Ch=1.5
Ch=0.5
0.00
2
Vs
3
4
0
1
2
Vs
3
Bright soil
Medium soil
Dark soil
Bright soil
Medium soil
Dark Soil
LAI=5
Vs=1
0.03
0.03
TCARI/OSAVI
TCARI/OSAVI
4
0.04
0.04
Vs=2
0.02
LAI=3
0.02
LAI=1
0.01
Vs=3
0.01
279
0.00
0
1
2
3
LAI
4
5
6
0
20
40
60
80
100
Cab
Fig. 6. Effects on TCARI/OSAVI for: i) vine width (C w) and height (C H) as function of visible soil strip (V s) and row LAI (top left and right); ii) visible soil strip (V s)
as function of row LAI and soil background (q s) (bottom left); and iii) row LAI and soil background (q s) on TCARI/OSAVI as function of C ab (bottom right).
280
0.40
ρ soil
0.30
0.20
0.10
0.00
400
500
600
700
Wavelength (nm)
800
900
Fig. 7. CASI soil spectra extracted from vineyard study sites and used as input
for the simulation methods.
5 –7% at 700 nm as function of the soil brightness levels. The
effects of row LAI, vineyard row width and height, visible soil
strip, and chlorophyll content were then studied for the TCARI/
OSAVI index, a combined ratio that has previously proven
successful in minimizing background effects for C ab retrieval.
The study of vine width and height on TCARI/OSAVI as
function of visible soil strip and row LAI (Fig. 6, top left and
right) indicates that larger effects are expected due to vine width
than to vine height, with larger effects on TCARI/OSAVI as
LAI increases. As the visible soil strip increases, vine width
effects decrease due to the greater contribution effects of soil
reflectance, with lower contribution of vegetation. As LAI
increases, large effects on TCARI/OSAVI are found when the
visible soil strip is small, i.e. close to 1 m (Fig. 6, bottom left).
Nevertheless, for visible soil strips greater than 1 m, i.e. 2 or 3 m
(an average of 2 m in the vine sites of this study) simulations
suggest a small effect of LAI variation. These simulations
indicate that larger effects are found on the index as function of
the visible soil strip (background effects) than due to the LAI
variation, suggesting the importance for properly describing the
vineyard architecture and soil characteristics. When LAI is
greater than 1, i.e. 3 or 5 (Fig. 6, bottom right) simulation results
suggest the low effects of LAI on TCARI/OSAVI as function of
C ab in vineyards with average dimensions (height = 1.5 m;
width = 1 m; visible soil strip = 2 m). In addition, in low LAI and
low C ab sites (Fig. 6, bottom left and right) soil backgrounds are
shown to be important, decreasing the effects on TCARI/
OSAVI as row LAI and C ab increase.
These simulations indicate the importance of properly
describing the vineyard architecture and dimensions, and the
soil background for accurate estimates of chlorophyll concentration. Therefore, in this study vineyard structural parameters
and airborne-sun viewing geometry angles that varied between
fields were taken into account as inputs for scaling-up methods
with the rowMCRM canopy model for simulation on each one of
the 103 study plots. Vineyard planting grids ranged between
2.5 1 and 3 1.5 m, vine height between 1.2 and 1.8 m, vine
width between 0.6 and 1.3 m, visible soil strip between 1.7 and
2.3 m, difference between sun azimuth and row direction
between 11- and 95-, and sun zenith angle between 31.5- and
46.7- for all images collected on the airborne campaigns for the 2
years and 103 sites. In addition, simulation methods for C ab
estimation employed the input of soil reflectance spectra
obtained directly from the imagery on areas of canopy openings
or missing vines within the field. Fig. 7 shows a range of soil
spectra extracted from the CASI imagery from each of the 14
fields acquired in the 2003 campaign, illustrating the gradient in
soil brightness and differences greater than 10% reflectance.
Predictive relationships were calculated for each field study
site between C ab and optical indices, using image and fieldmeasured parameters for the vineyard structure, soil background and viewing geometry. Scaling-up methods used here
are similar to the ones described in Zarco-Tejada et al. (2001)
for forest canopies, Haboudane et al. (2002, 2004) for corn
crops using PROSPECT-SAILH, and Zarco-Tejada et al.
(2004) for open tree canopy crops using PROSPECT-SAILHFLIM models. Scaling-up relationships were developed for
each study site using a range of LAI values between 1 and 5,
and between 5 and 95 Ag/cm2 for C ab, fixing the remaining leaf
0.40
120
y = 188.29e-12.588x
R2 = 0.9671
100
0.30
ρsoil
Cab
80
60
0.20
40
Bright soil
Medium soil
Dark soil
20
0
0
0.1
0.2
0.3
Modeled TCARI/OSAVI
(PROSPECT-row MCRM)
0.4
0.10
Bright soil
Medium soil
Dark soil
0.00
400
500
600
700
800
900
Wavelength (nm)
Fig. 8. Example of 3 predictive relationships (left) developed for study sites with extreme soil backgrounds (right). Relationships were developed using a range of
row LAI values between 1 and 5, C ab ranging between 5 and 95 Ag/cm2 and structural parameters measured in the field. The remaining parameters were fixed to the
values shown in Table 5.
parameters to the values estimated by model inversion from the
leaf samples previously described, with the field structural
measurements describing the vineyard canopy structure on
each study site (Table 5; Fig. 8). Estimated C ab for each of the
103 study sites was then compared with the field-measured
values of pigment content obtained by destructive sampling
and described in the previous section.
281
measured are presented in this section. Results for the
regression analysis between leaf reflectance indices and
biochemical constituents, leaf model inversion with PROSPECT on vine leaves, and C ab estimation at canopy level with
PROSPECT – rowMCRM models through scaling-up methods
are described.
5. Results
5.1. Relationships between optical indices and C ab for V.
vinifera L. leaves
Analysis conducted on the 605 leaves where optical
properties and leaf biochemical constituents C a, C b, C ab,
C x+c and ratios C a / C b, C a / C x+c , C b / C x+c , C ab / C x+c were
Results of the relationships found between 45 narrow-band
optical indices calculated from leaf reflectance spectra (Table
4), and biochemical constituents C a, C b, C ab, C x+c and ratios
Table 6
Determination coefficients (r 2) found for linear relationships (left on each cell), exponential (center) and 3rd order polynomial (right) between leaf optical indices and
biochemical constituents (n = 605)
Ca
NDVI
RDVI
SR
MSR
G
ZM
VOG1
VOG2
VOG3
GM1
GM2
CTR1
CTR2
RGI
BGI1
BGI2
BRI1
BRI2
CUR
LIC1
LIC2
LIC3
SIPI
PRI1
PRI2
PRI3
NPCI
SRPI
NPQI
MCARI
TCARI
OSAVI
MCARI1
MCARI2
MTVI1
MTVI2
TVI
MSAVI
MCARI/OSAVI
TCARI/OSAVI
MCARI1/OSAVI
MCARI2/OSAVI
MTVI1/OSAVI
MTVI2/OSAVI
TVI/OSAVI
0.26
0.34
0.36
0.34
0.53
0.89
0.89
0.87
0.86
0.87
0.81
0.16
0.41
0.35
0.10
0.77
0.01
0.35
0.33
0.23
0.25
0.08
0.36
0.41
0.13
0.35
0.04
0.03
0.01
0.66
0.74
0.29
0.23
0.23
0.02
0.00
0.01
0.36
0.68
0.70
0.16
0.09
0.44
0.60
0.53
Cb
0.52 0.37
0.60 0.56
0.56 0.37
0.58 0.36
0.32 0.60
0.83 0.89
0.84 0.89
0.76 0.87
0.75 0.87
0.82 0.88
0.85 0.82
0.25 0.22
0.71 0.65
0.12 0.54
0.10 0.14
0.64 0.77
0.02 0.02
0.53 0.36
0.41 0.33
0.48 0.33
0.39 0.28
0.21 0.10
0.62 0.43
0.34 0.44
0.02 0.17
0.44 0.36
0.11 0.05
0.09 0.05
0.01 0.01
0.66 0.79
0.73 0.83
0.56 0.47
0.44 0.31
0.46 0.34
0.12 0.09
0.02 0.01
0.04 0.21
0.62 0.54
0.82 0.87
0.90 0.89
0.36 0.18
0.15 0.12
0.66 0.50
0.70 0.67
0.70 0.62
Highlighted are results for r 2 > 0.7.
0.23
0.31
0.34
0.32
0.50
0.84
0.84
0.82
0.82
0.84
0.77
0.14
0.37
0.35
0.09
0.72
0.00
0.31
0.30
0.21
0.22
0.08
0.34
0.40
0.13
0.34
0.03
0.02
0.00
0.61
0.69
0.27
0.20
0.20
0.01
0.00
0.01
0.33
0.62
0.64
0.14
0.09
0.41
0.56
0.49
C ab
0.43
0.50
0.52
0.52
0.36
0.82
0.82
0.76
0.75
0.82
0.82
0.22
0.61
0.17
0.09
0.64
0.02
0.46
0.40
0.38
0.34
0.17
0.54
0.37
0.04
0.44
0.08
0.06
0.01
0.67
0.74
0.46
0.35
0.37
0.07
0.00
0.01
0.53
0.78
0.83
0.29
0.14
0.60
0.69
0.66
0.35
0.54
0.35
0.35
0.57
0.84
0.84
0.82
0.83
0.84
0.78
0.20
0.61
0.53
0.13
0.72
0.01
0.31
0.30
0.31
0.24
0.09
0.42
0.42
0.16
0.34
0.04
0.04
0.01
0.74
0.78
0.45
0.29
0.31
0.07
0.08
0.19
0.52
0.81
0.83
0.16
0.11
0.48
0.64
0.60
0.26
0.33
0.36
0.34
0.53
0.89
0.89
0.87
0.87
0.88
0.81
0.16
0.41
0.35
0.09
0.77
0.01
0.35
0.33
0.23
0.25
0.08
0.36
0.41
0.13
0.36
0.03
0.03
0.01
0.65
0.74
0.29
0.22
0.22
0.02
0.00
0.01
0.36
0.68
0.69
0.15
0.10
0.44
0.60
0.53
C x+c
0.50 0.37
0.58 0.57
0.56 0.37
0.57 0.36
0.34 0.60
0.84 0.89
0.85 0.89
0.78 0.87
0.76 0.87
0.84 0.88
0.85 0.82
0.25 0.22
0.69 0.65
0.14 0.55
0.10 0.14
0.65 0.77
0.02 0.02
0.52 0.35
0.41 0.33
0.46 0.33
0.38 0.27
0.20 0.10
0.61 0.44
0.36 0.44
0.03 0.17
0.45 0.36
0.11 0.05
0.08 0.05
0.01 0.00
0.68 0.79
0.75 0.83
0.54 0.47
0.41 0.31
0.44 0.34
0.11 0.08
0.01 0.02
0.03 0.20
0.61 0.54
0.82 0.86
0.90 0.89
0.34 0.18
0.15 0.12
0.65 0.50
0.71 0.67
0.70 0.62
0.25
0.33
0.31
0.30
0.44
0.76
0.76
0.73
0.73
0.76
0.69
0.15
0.38
0.27
0.08
0.66
0.00
0.33
0.27
0.23
0.23
0.08
0.33
0.25
0.06
0.23
0.03
0.02
0.01
0.54
0.61
0.29
0.25
0.25
0.04
0.01
0.00
0.35
0.58
0.62
0.13
0.06
0.37
0.49
0.45
0.41 0.31
0.50 0.51
0.42 0.31
0.44 0.31
0.32 0.51
0.74 0.76
0.74 0.76
0.69 0.73
0.68 0.74
0.74 0.76
0.73 0.70
0.20 0.20
0.56 0.55
0.14 0.48
0.08 0.13
0.59 0.66
0.01 0.01
0.42 0.33
0.31 0.28
0.38 0.29
0.31 0.26
0.17 0.10
0.48 0.37
0.22 0.29
0.01 0.11
0.27 0.23
0.08 0.06
0.06 0.06
0.01 0.00
0.54 0.66
0.60 0.70
0.45 0.41
0.39 0.32
0.39 0.34
0.11 0.10
0.02 0.07
0.03 0.19
0.51 0.48
0.65 0.74
0.73 0.77
0.24 0.17
0.08 0.09
0.50 0.40
0.55 0.53
0.54 0.50
Ca / Cb
C a / C x+c
0.05
0.05
0.01
0.02
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.05
0.02
0.00
0.00
0.00
0.03
0.00
0.05
0.02
0.02
0.03
0.01
0.01
0.01
0.03
0.02
0.01
0.00
0.01
0.05
0.05
0.05
0.05
0.05
0.05
0.04
0.01
0.02
0.03
0.00
0.02
0.00
0.01
0.37
0.40
0.46
0.46
0.27
0.68
0.69
0.63
0.62
0.66
0.70
0.19
0.52
0.12
0.08
0.51
0.02
0.40
0.36
0.32
0.30
0.14
0.47
0.43
0.07
0.49
0.09
0.07
0.01
0.60
0.65
0.38
0.25
0.27
0.04
0.00
0.00
0.43
0.70
0.72
0.30
0.18
0.56
0.62
0.59
0.09 0.08
0.09 0.08
0.02 0.07
0.04 0.08
0.00 0.04
0.00 0.04
0.01 0.05
0.00 0.03
0.00 0.03
0.00 0.05
0.01 0.05
0.02 0.02
0.09 0.09
0.03 0.06
0.000 0.00
0.00 0.00
0.00 0.00
0.05 0.05
0.00 0.01
0.10 0.08
0.04 0.06
0.04 0.08
0.06 0.08
0.01 0.00
0.01 0.05
0.00 0.01
0.04 0.05
0.03 0.05
0.00 0.01
0.00 0.02
0.00 0.01
0.09 0.08
0.09 0.08
0.09 0.08
0.09 0.08
0.08 0.08
0.09 0.08
0.08 0.08
0.02 0.02
0.04 0.06
0.06 0.07
0.00 0.01
0.04 0.05
0.01 0.03
0.02 0.04
0.48
0.50
0.53
0.54
0.20
0.64
0.65
0.57
0.56
0.62
0.68
0.22
0.64
0.06
0.08
0.46
0.03
0.46
0.39
0.42
0.35
0.19
0.56
0.39
0.03
0.51
0.12
0.10
0.01
0.57
0.61
0.49
0.33
0.37
0.09
0.02
0.03
0.53
0.73
0.78
0.38
0.18
0.63
0.64
0.64
0.48
0.53
0.49
0.48
0.39
0.75
0.75
0.74
0.74
0.73
0.73
0.25
0.66
0.29
0.10
0.58
0.03
0.44
0.38
0.43
0.37
0.17
0.51
0.43
0.13
0.50
0.11
0.11
0.01
0.61
0.65
0.52
0.28
0.32
0.11
0.06
0.18
0.53
0.72
0.74
0.31
0.19
0.59
0.64
0.63
C b / C x+c
C ab / C x+c
0.18
0.20
0.29
0.26
0.23
0.45
0.46
0.43
0.42
0.45
0.46
0.09
0.27
0.14
0.05
0.35
0.01
0.21
0.24
0.15
0.15
0.06
0.26
0.32
0.07
0.34
0.02
0.02
0.01
0.40
0.45
0.19
0.10
0.12
0.00
0.00
0.01
0.22
0.44
0.43
0.15
0.12
0.33
0.42
0.37
0.33
0.36
0.44
0.42
0.28
0.65
0.66
0.60
0.59
0.63
0.66
0.17
0.47
0.14
0.07
0.49
0.02
0.36
0.35
0.28
0.27
0.12
0.43
0.43
0.07
0.48
0.07
0.06
0.01
0.57
0.62
0.34
0.21
0.24
0.03
0.01
0.00
0.39
0.66
0.67
0.27
0.17
0.52
0.60
0.56
0.22 0.31
0.24 0.35
0.33 0.31
0.31 0.31
0.20 0.26
0.46 0.47
0.46 0.47
0.43 0.46
0.42 0.47
0.45 0.47
0.48 0.46
0.12 0.14
0.33 0.42
0.11 0.20
0.05 0.06
0.35 0.39
0.01 0.01
0.25 0.23
0.28 0.24
0.18 0.27
0.19 0.19
0.08 0.08
0.31 0.34
0.34 0.32
0.06 0.09
0.38 0.34
0.04 0.03
0.03 0.03
0.01 0.01
0.43 0.43
0.48 0.45
0.23 0.34
0.13 0.16
0.15 0.19
0.01 0.03
0.01 0.04
0.01 0.09
0.27 0.35
0.49 0.46
0.49 0.47
0.18 0.17
0.12 0.12
0.38 0.38
0.46 0.44
0.42 0.42
0.41 0.45
0.44 0.50
0.50 0.46
0.50 0.45
0.21 0.37
0.62 0.71
0.63 0.71
0.56 0.69
0.55 0.69
0.61 0.69
0.66 0.69
0.20 0.23
0.57 0.62
0.08 0.28
0.07 0.10
0.45 0.56
0.02 0.02
0.42 0.39
0.38 0.36
0.36 0.41
0.32 0.33
0.16 0.15
0.51 0.49
0.40 0.43
0.04 0.12
0.50 0.48
0.09 0.08
0.08 0.08
0.01 0.01
0.57 0.59
0.62 0.62
0.43 0.49
0.28 0.25
0.31 0.30
0.06 0.09
0.001 0.03
0.01 0.16
0.47 0.50
0.70 0.68
0.74 0.70
0.33 0.28
0.18 0.18
0.58 0.56
0.63 0.62
0.61 0.60
282
are shown in Table 6. The determination coefficients obtained
for regression analysis using linear, exponential and 3rd order
polynomials demonstrate the non-linearity between specific
indices and the biochemical constituents. Red edge ratio
indices are generally linear, such as the Vogelmann (VOG2)
or ZM index, yielding r 2 = 0.87 and r 2 = 0.89, respectively for
C ab (Fig. 9, top). Other indices showed a clear non-linear
relationship with C ab, such as those calculated from visible
bands only or in addition to red edge wavelengths (OSAVI,
MCARI1, MCARI2, MSAVI, PRI3), and those indices calculated as combined with OSAVI (MCARI/OSAVI, TCARI/
OSAVI, etc) (Fig. 9, center for PRI3 and TCARI/OSAVI). The
best optical indices for correlation with C ab in V. vinifera L.
leaves were ZM (r 2 = 0.89, linear), VOG1 (r 2 = 0.89, linear),
VOG2 (r 2 = 0.87, linear), VOG3 (r 2 = 0.87, linear), GM1
(r 2 = 0.88, linear), GM 2 (r 2 = 0.85, exponential), BGI 2
(r 2 = 0.77, linear), MCARI (r 2 = 0.79, 3rd order polynomial),
TCARI (r 2 = 0.83, 3rd order polynomial), MCARI/OSAVI
(r 2 = 0.86, 3rd order polynomial), and TCARI/OSAVI
(r 2 = 0.9, exponential). As expected, indices traditionally used
for vegetation monitoring, such as NDVI, SR or MSR did not
obtain as good results as red edge and combined indices,
80
100
y = -282.74x + 10.116
r2 = 0.87
60
Cab (µg/cm2)
Cab (µg/cm2)
80
60
40
40
20
20
y = 22.87x - 18.248
r2 = 0.89
0
0
-0.3
-0.2
-0.1
0
0
y = 33.562e-18.126x
r2 = 0.45
3
4
5
y = 63.541e-2.1792x
r2 = 0.9
60
Cab (µg/cm2)
60
Cab (µg/cm2)
2
80
80
40
40
20
20
0
-0.05
16
1
Z & M (R750) / (R710)
VOG2 (R734 – R747)/(R715 + R726)
0
0
0.05
PRI3 (R570 – R539)/(R570 + R539)
0.1
0
0.5
1
TCARI / OSAVI
1.5
7
y = 0.4963e3.6243x
r2 = 0.49
y = 4.2206e-9.1789x
r2 = 0.5
5
Cab / Cx+c
Cx+c (µg/cm2)
12
8
3
4
0
0.25
0.5
0.75
SIPI = (R800 – R450)/(R800 + R650)
1
1
-0.05
0
0.05
PRI3 (R570 – R539)/(R570 + R539)
0.1
Fig. 9. Relationships obtained between C ab and optical indices calculated from leaf reflectance spectra VOG2 (upper left) and ZM (upper right), showing a non-linear
behavior with PRI3 (center left) and TCARI/OSAVI (center right), SIPI with C x+c (bottom left) and PRI3 with C ab / C x+c (bottom right).
0.6
Measured R
Measured T
5.2. Estimation of C ab by PROSPECT inversion in V. vinifera
L. leaves
Modeled R
Modeled T
R
R&T
0.4
T
0.2
0
400
500
600
700
800
Wavelength (nm)
80
Cab (µg/cm2) (measured)
283
60
40
The V. vinifera L. leaf reflectance and transmittance
database used for inversion with PROSPECT yielded a good
agreement with the modeled spectra, obtaining an average
RMSE = 0.025 for the 605 leaves (Fig. 10, top). The input
variables N, C ab and C m for PROSPECT, estimated for each
leaf spectra by inversion using the iterative optimization
method between 400 and 800 nm, yielded average values for
the 605 leaves for N (l = 1.62, r = 0.14), C ab (l = 39.4,
r = 13.4), and C m (l = 0.0035, r = 0.0012). The relationship
between the measured C ab for each vine leaf, and the
PROSPECT-inverted C ab from the optical measurements on
the same leaves yielded a determination coefficient of r 2 = 0.95
and RMSE = 5.3 Ag/cm2 (Fig. 10, bottom). The PROSPECT
model was shown to be valid for simulating the leaf optical
properties of V. vinifera L. leaves, although a slight overestimation of C ab was found when compared to the 1 : 1
relationship (yielding the RMSE = 5.3 Ag/cm2 mentioned).
The RMSE obtained is within the normal range of variation
found in similar studies with other species that, in conjunction
with the high determination coefficient obtained for this large
database, demonstrates the applicability of PROSPECT to
simulate the optical properties of V. vinifera L. leaves.
20
y = 0.937x - 2.6738
r2 = 0.95
RMSE=5.3 µg/cm2
0
0
20
40
60
80
Cab (µg/cm2) (estimated)
Fig. 10. Leaf reflectance and transmittance spectra measured with the Li-Cor
1800-12 integrating sphere and simulated with PROSPECT (top). Relationship
obtained between the C ab measured by destructive sampling and estimated by
PROSPECT inversion using the leaf optical properties (bottom).
yielding r 2 = 0.5 (NDVI, exponential), r 2 = 0.56 (SR, exponential), and r 2 = 0.57 (MSR, exponential). Indices developed for
maximizing its sensitivity to LAI while decreasing C ab effects,
such as MCARI2 and MTVI2 (Haboudane et al., 2004),
demonstrated a low relationship with C ab, as expected. Among
these structural indices that are demonstrated to be highly
related to LAI, MTVI2 was shown in this study to be less
affected by C ab variations (r 2 = 0.02) than the MCARI2 index
(r 2 = 0.44, exponential).
Regression results for indices and C x+c and ratios C ab / C x+c
generally showed poorer relationships, obtaining r 2 = 0.49 for
C x+c with SIPI, and r 2 = 0.5 for C ab / C x+c with PRI3 (Fig. 9,
bottom). The PRI index developed for xanthophyll cycle
pigment change detection (Gamon et al., 1992) was shown to
be more related to the C ab / C x+c ratio (r 2 = 0.5, exponential)
than to C ab alone (r 2 = 0.45) and C x+c (r 2 = 0.27). This result
agrees with Sims and Gamon (2002) who suggested PRI as a
potential indicator for carotenoid / chlorophyll ratio monitoring.
With respect to chlorophyll a and b ratios, none of the indices
proposed were related to the chlorophyll ratio C a / C b, all
yielding poor results.
5.3. Estimation of C ab by PROSPECT – rowMCRM in V.
vinifera L. fields
The narrow-band indices that obtained the best relationships
in the leaf-level study for C ab estimation, plus the traditional
index NDVI (Table 6), were calculated from the 103 sites of
10 10 m imaged by ROSIS and CASI sensors. Relationships
were obtained between field-measured C ab and the indices
calculated from the airborne reflectance for all pixels falling
within the 10 10 m site (pure vine + soil + shadows) and for
Table 7
Determination coefficients (r 2) obtained between ROSIS and CASI airborne
optical indices and C ab for the 103 study sites imaged
Indices calculated from
ROSIS and CASI images
NDVI
ZM
VOG1
VOG2
VOG3
GM1
GM2
CTR2
MCARI
TCARI
MCARI/OSAVI
TCARI/OSAVI
MTVI2/OSAVI
TVI/OSAVI
Highlighted are results for r 2 > 0.5.
Chlorophyll content (C ab)
All pixels
(soil + vegetation)
Pure vegetation
pixels
0.00
0.00
0.00
0.03
0.03
0.11
0.00
0.21
0.40
0.43
0.61
0.59
0.25
0.23
0.36
0.24
0.25
0.31
0.30
0.07
0.21
0.14
0.54
0.58
0.53
0.55
0.51
0.49
284
100
100
y = -451.25x + 58.832
r2 = 0.43
80
y = 65.6e-4.2101x
r2 = 0.59
80
60
Cab
Cab
60
40
40
20
20
0
0
0
0.02
0.04
0.06
0.08
0
0.1
TCARI (all pixels)
0.2
0.3
TCARI/OSAVI (all pixels)
100
100
y = -310.39x + 61.055
r2 = 0.58
80
y = 71.901e-4.6579x
r2 = 0.55
80
Cab
60
Cab
60
40
40
20
20
0
0
0
0.05
0.1
0.15
0
0.1
TCARI (veg. pixels)
0.2
0.3
0.4
TCARI/OSAVI (veg. pixels)
Fig. 11. Relationships obtained between C ab and ROSIS and CASI indices TCARI (top and bottom left), and TCARI/OSAVI (top and bottom right) for aggregated
pixels (top left and right) and pure vine pixels (bottom left and right).
aggregated pixels and r 2 = 0.36 on pure vine pixels, demonstrating that it is not appropriate for vineyard condition
monitoring on non-homogeneous canopies imaged with spatial
resolutions lower than 1 m pixel size due to the large
background effects and low sensitivity to pigment concentration as indicator of physiological status.
100
Cab (µg/cm2) (measured)
the pure vine reflectance only at each study site as extracted
with the high-resolution imagery (Table 7). Consistent with
previous studies in non-homogeneous crop canopies (ZarcoTejada et al., 2004), MCARI, TCARI, and combined indices
MCARI/OSAVI and TCARI/OSAVI indices yielded the best
relationships for both aggregated and pure vegetation pixels.
MCARI (TCARI) yielded r 2 = 0.40 (r 2 = 0.43) for aggregated
pixels, and r 2 = 0.54 (r 2 = 0.58) for pure vine pixels, showing
the effects of soil and shadows on both indices (Fig. 11, top and
bottom left for TCARI). Combined indices MCARI/OSAVI
and TCARI/OSAVI showed less sensitivity to background
effects, as expected, yielding r 2 = 0.61 and r 2 = 0.59 for
aggregated pixels, respectively (Fig. 11 top and bottom right
for TCARI/OSAVI). The TCARI/OSAVI index showed the
greatest consistency when calculated for aggregated and pure
vine pixels (r 2 = 0.59 for aggregated pixels; r 2 = 0.55 for pure
vine pixels), suggesting this as the most robust narrow-band
index for vineyard pigment content monitoring. Other vegetation indices that show significant results at the leaf-level, such
as ZM (r 2 = 0.89 at leaf-level), VOG1, 2, 3 (r 2 = 0.8), GM1, 2
(r 2 = 0.8), and CTR2 (r 2 = 0.69), were shown to be totally
unsuccessful when applied to image-level aggregated pixels
due to their high sensitivity to soil background (r 2¨0.1),
generating a maximum of r 2¨0.3 when applied to pure vine
pixels. The traditional NDVI index, generally used for
vegetation biomass and vigor monitoring, yielded r 2¨0 on
80
60
40
y = 1.2444x - 9.7719
r2 = 0.67
RMSE=11.5 µg/cm2
20
0
0
20
Cab
40
(µg/cm2)
60
80
100
(estimated)
Fig. 12. Estimation of vine C ab at the canopy level by scaling-up TCARI/
OSAVI through PROSPECT linked to rowMCRM model.
Fig. 13. Airborne CASI image of 1 m spatial resolution and 8 spectral bands
(top) showing C ab estimated with TCARI/OSAVI index through PROSPECT –
rowMCRM linked models (bottom).
Prediction relationships obtained with PROSPECT –
rowMCRM models as explained in the Methods section (Table
5; Fig. 8) when applied to the 103 study sites imaged by ROSIS
and CASI airborne sensors yielded r 2 = 0.67 (RMSE = 11.5 Ag/
cm2) for C ab estimation (Fig. 12). These results at leaf, canopy
level for specific indices, and through scaling-up methods,
suggest the successful retrieval capability of C ab in rowstructured vineyard canopies. An example of the natural
variability detected in C ab at the vine level in different fields
imaged by the CASI sensor in July 2003 can be seen in Fig. 13,
illustrating the C ab product in 5 ranges of chlorophyll
concentration. The provision of a C ab product map in 5 steps
is considered consistent with the RMSE = 11.5 Ag/cm2 retrieval
accuracy expected.
6. Conclusions
This study investigated the optical properties of V. vinifera
L. leaves through reflectance and transmittance measurements,
optical index calculation, and destructive determination of
pigments using a large database of 1467 leaves collected in
summer 2002 and 2003. Airborne campaigns imaged a total of
103 study sites from 24 vineyard fields with the ROSIS and
285
CASI hyperspectral sensors at 1 m spatial resolution, studying
the validity of optical indices generally used with success in
other species at leaf and canopy levels through scaling-up
simulation with PROSPECT and rowMCRM row-structured
canopy reflectance model.
A measurement protocol using a Li-Cor 1800-12 integrating
sphere attached to an Ocean Optics model USB2000 fiber
spectrometer for stray-light corrected reflectance and transmittance measurements was presented. The measurement protocol
consisted of a total of five measurements modifying the
position of the collimated light, dark and white plugs of the
integrating sphere to measure the reflectance and transmittance
signals, reflectance internal standard, reflectance ambient, and
dark measurement. The best optical indices for correlation with
C ab in V. vinifera L. leaves were ZM, VOG1, VOG2, VOG3,
GM1, GM2, BGI2, MCARI, TCARI, MCARI/OSAVI, and
TCARI/OSAVI (r 2 ranging between 0.8 and 0.9), with poor
performance of traditional indices NDVI, SR or MSR. Linear
relationships were found between red edge ratio indices such as
ZM and VOG indices, whereas generally non-linear relationships were derived with combined indices and ratio indices
with visible bands. Results for C x+c and C ab / C x+c ratios
yielded r 2 = 0.49 for C x+c with SIPI, and r 2 = 0.5 for C ab / C x+c
with PRI3. The PRI index was shown as a potential indicator
for carotenoid / chlorophyll ratio monitoring.
The inversion of PROSPECT model for N, C ab, C m and C w
estimation, using the large subset database of 605 vine leaf
spectra, obtained an averaged RMSE of 0.025, yielding mean
values of N = 1.62, C ab = 39.4, C w = 0.02, and C m = 0.0035
(r 2 = 0.95 and RMSE = 5.3 Ag/cm2 for C ab estimation by
inversion). Therefore these results demonstrate that the PROSPECT leaf model is valid for simulation of the optical properties
of vine leaves as function of different pigment levels.
The leaf-level indices that produced the best correlations
with C ab were tested at the canopy level on vineyard
reflectance spectra extracted from CASI and ROSIS hyperspectral images collected from 103 sites in 24 fields over 2
years. Results at the canopy level demonstrated that MCARI,
TCARI, and combined indices MCARI/OSAVI and TCARI/
OSAVI indices generated the best relationships for both
aggregated and pure vegetation pixels. Combined indices
MCARI/OSAVI and TCARI/OSAVI showed less sensitivity
to background effects, yielding r 2 = 0.61 and r 2 = 0.59 for
aggregated pixels containing pure vine, soil and shadow
components. TCARI/OSAVI was the most consistent index
for estimating C ab on aggregated and pure vine pixels, yielding
r 2 = 0.59 for aggregated pixels and r 2 = 0.55 for pure vine
pixels. Physical methods based on PROSPECT linked to
rowMCRM model enabled accounting for vineyard structure,
row orientation, viewing geometry and background effects,
indicating the large effects of the background and vine
dimensions on the canopy reflectance. Predictive relationships
were developed using PROSPECT – rowMCRM model between C ab and TCARI/OSAVI as function of LAI, using fieldmeasured vine dimensions and image-extracted soil background, row-orientation and viewing geometry. Model-based
prediction relationships for C ab content were successfully
286
applied to the 103 study sites imaged by ROSIS and CASI
airborne sensors, yielding r 2 = 0.67 (RMSE = 11.5 Ag/cm2).
Results presented in this manuscript indicate the validity of
narrow-band indices for C ab estimation and chlorosis detection
at the leaf and canopy levels in V. vinifera L., demonstrating
the validity of PROSPECT and rowMCRM models for leaf and
canopy level estimations. This methodology for scaling-up
leaf-level sensitive indices enabled conclusions to be reached
on the effectiveness of biochemical constituent retrievals in
canopies where model inversions are complex due to the large
number of input parameters required to feed the linked leafcanopy model.
Acknowledgments
The authors gratefully acknowledge the HySens HS2002-E1
project support provided through the Access to Research
Infrastructures EU Program. Financial support from the
Spanish Ministry of Science and Technology (MCyT) for the
project AGL2002-04407-C03, and financial support to P.J.
Zarco-Tejada under the Ramón y Cajal and Averroes Programs
are also acknowledged. Financial support from the Natural
Sciences and Engineering Research Council (NSERC) of
Canada to permit contributions by J.R. Miller is gratefully
acknowledged. We thank S. Holzwarth, A. Müeller and the rest
of the DLR group, L. Gray and J. Freemantle from York
University, and the INTA group for efficient airborne field
campaigns with CASI, ROSIS, and DAIS sensors, providing
coordination with field data collection. Fertiberia S.A. is
acknowledged for providing determinations of foliar nutrients.
A. Kuusk and J. Praks are gratefully acknowledged for sharing
the rowMCRM code, and E. Vera-Toscano for providing
valuable suggestions.
References
Allen, W. A., Gayle, & Richarson, A. J. (1970). Plant canopy irradiance
specified by the Duntley equations. Journal of the Optical Society of
America, 60(3), 372 – 376.
Allen, W. A., Gausman, H. W., Richarson, A. J., & Thomas, J. R. (1969).
Interaction of isotropic light with compact plan leaf. Journal of the Optical
Society of America, 59(10), 1376 – 1379.
Barnes, J. D. (1992). A reappraisal of the use of DMSO for the extraction and
determination of chlorophylls a and b in lichens and higher plants.
Environmental Experimental Botany, 2, 85 – 100.
Ben-Dor, E., & Levin, N. (2000). Determination of surface reflectance from
raw hyperspectral data without simultaneous ground data measurements: A
case study of the GER 63-channel sensor data acquired over Naan, Israel.
International Journal of Remote Sensing, 21, 2053 – 2074.
Broge, N. H., & Leblanc, E. (2000). Comparing prediction power and stability
of broadband and hyperspectral vegetation indices for estimation of green
leaf area index and canopy chlorophyll density. Remote Sensing of
Environment, 76, 156 – 172.
Carter, G. A. (1991). Primary and secondary effects of water content of the
spectral reflectance of leaves. American Journal of Botany, 78, 916 – 924.
Carter, G. A. (1994). Ratios of leaf reflectances in narrow wavebands as
indicators of plant stress. International Journal of Remote Sensing, 15,
697 – 704.
Carter, G. A., Dell, T. R., & Cibula, W. G. (1996). Spectral reflectance
characteristics and digital imagery of a pine needle blight in the southeastern
United States. Canadian Journal of Forest Research, 26, 402 – 407.
Carter, G. A., & Spiering, B. A. (2002). Optical properties of intact leaves for
estimating chlorophyll concentration. Journal of Environmental Quality,
31(5), 1424 – 1432.
Ceccato, P., Flasse, S., Tarantola, S., Jacquemoud, S., & Gregoire, J. M. (2001).
Detecting vegetation leaf water content using reflectance in the optical
domain. Remote Sensing of Environment, 77, 22 – 33.
Chen, J. (1996). Evaluation of vegetation indices and modified simple ratio for
boreal applications. Canadian Journal of Remote Sensing, 22, 229 – 242.
Danson, F. M., Steven, M. D., Malthus, T. J., & Clarck, J. A. (1992). Highspectral resolution data for determining leaf water content. International
Journal of Remote Sensing, 13, 461 – 470.
Daughtry, C. S. T., Walthall, C. L., Kim, M. S., Brown de Colstoun, E., &
McMurtrey, J. E. III (2000). Estimating corn leaf chlorophyll concentration
from leaf and canopy reflectance. Remote Sensing of the Environment, 74,
229 – 239.
Dobrowski, S. Z., Ustin, S. L., & Wolpert, J. A. (2002). Remote estimation of
vine canopy density in vertically shoot-positioned vineyards: determining
optimal vegetation indices. Australian Journal of Grape and Wine
Research, 8(2), 117 – 125.
Dobrowski, S. Z., Ustin, S. L., & Wolpert, J. A. (2003). Grapevine dormant
pruning weight prediction using remotely sensed data. Australian Journal
of Grape and Wine Research, 9(3), 177 – 182.
Fernandez-Escobar, R., Moreno, R., & Garcia-Creus, M. (1999). Seasonal
changes of mineral nutrients in olive leaves during the alternate-bearing
cycle. Scientia Horticulturae, 82, 24 – 45.
Fourty, T., & Baret, F. (1997). Vegetation water and dry matter contents
estimated from top-of-the-atmosphere reflectance data: A simulation study.
Remote Sensing of Environment, 61(1), 34 – 45.
Fuentes, D. A., Gamon, J. A., Qiu, H., Sims, D. A., & Roberts, D. A. (2001).
Mapping Canadian boreal forest vegetation using pigment and water
absorption features derived from AVIRIS sensor. Journal of Geophysical
Research, 106, 33565 – 33577.
Gamon, J. A., Peñuelas, J., & Field, C. B. (1992). A narrow-waveband spectral
index that tracks diurnal changes in photosynthetic efficiency. Remote
Sensing of Environment, 41, 35 – 44.
Gamon, J. A., & Surfus, J. S. (1999). Assessing leaf pigment content and
activity with a reflectometer. New Phytologist, 143, 105 – 117.
Gao, B.-C. (1996). NDWI — a normalized difference water index for remote
sensing of vegetation liquid water from space. Remote Sensing of
Environment, 58, 257 – 266.
Gitelson, A. A., Gritz, Y., & Merzlyak, M. N. (2003). Relationships between
leaf chlorophyll content and spectral reflectance and algorithms for nondestructive chlorophyll assessment in higher plants. Journal of Plant
Physiology, 160(3), 271 – 282.
Gitelson, A. A., & Merzlyak, M. N. (1996). Signature analysis of leaf
reflectance spectra: Algorithm development for remote sensing of chlorophyll. Journal of Plant Physiology, 148, 494 – 500.
Gitelson, A. A., & Merzlyak, M. N. (1997). Remote estimation of chlorophyll
content in higher plant leaves. International Journal of Remote Sensing, 18,
2691 – 2697.
Haboudane, D., Miller, J. R., Pattey, E., Zarco-Tejada, P. J., & Strachan, I.
(2004). Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of
precision agriculture. Remote Sensing of Environment, 90(3), 337 – 352.
Haboudane, D., Miller, J. R., Tremblay, N., Zarco-Tejada, P. J., & Dextraze, L.
(2002). Integration of hyperspectral vegetation indices for prediction of
crop chlorophyll content for application to precision agriculture. Remote
Sensing of Environment, 81(2 – 3), 416 – 426.
Hall, A., Lamb, D. W., Holzapfel, B., & Louis, J. (2002). Optical remote
sensing applications in viticulture — A review. Australian Journal of Grape
and Wine Research, 8(1), 36 – 47.
Hall, A., Louis, J., & Lamb, D. (2003). Characterising and mapping vineyard
canopy using high-spatial-resolution aerial multispectral images. Computers and Geosciences, 29, 813 – 822.
Harron, J. (2000). Optical properties of phytoelements in conifers. MSc. Thesis,
York University, Toronto, Canada.
Horler, D. N. H., Dockray, M., & Barber, J. (1983). The red edge of plant leaf
reflectance. International Journal of Remote Sensing, 4(2), 273 – 288.
Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing
of the Environment, 25, 295 – 309.
Jacquemoud, S., & Baret, F. (1990). Prospect: A model of leaf optical
properties spectra. Remote Sensing of Environment, 34, 75 – 91.
Jacquemoud, S., Ustin, S. L., Verdebout, J., Schmuck, G., Andreoli, G., &
Hosgood, B. (1996). Estimating leaf biochemistry using the PROSPECT leaf optical properties model. Remote Sensing of Environment,
56, 194 – 202.
Johnson, L. F. (2003). Temporal stability of an NDVI – LAI relationship in a
Napa Valley vineyard. Australian Journal of Grape and Wine Research,
9(2), 96 – 101.
Johnson, L. F., Bosch, D. F., Williams, D. C., & Lobitz, B. M. (2001). Remote
sensing of vineyard management zones: Implications for wine quality.
Applied Engineering in Agriculture, 17(4), 557 – 560.
Johnson, L. F., Roczen, D. E., Youkhana, S. K., Nemani, R. R., & Bosch, D. F.
(2003). Mapping vineyard leaf area with multispectral satellite imagery.
Computers and Electronics in Agriculture, 38, 33 – 44.
Jolley, V. D, & Brown, J. C. (1994). Genetically controlled uptake and use of
iron by plants. In J. A. Manthey, D. E. Crowley, & D. G. Luster (Eds.),
Biochemistry of metal micronutrients in the rhizosphere (pp. 251 – 266).
Boca Raton’ Lewis Publishers.
Jordan, C. F. (1969). Derivation of leaf area index from quality of light on the
forest floor. Ecology, 50, 663 – 666.
Kim, M. S., Daughtry, C. S. T., Chappelle, E. W., McMurtrey III, J. E., &
Walthall, C. L. (1994). The use of high spectral resolution bands for
estimating absorbed photosynthetically active radiation (Apar). 6th symp.
on physical measurements and signatures in remote sensing. Jan. 17 – 21,
1994, Val D’Isere, France.
Kuusk, A. (1995). A Markov chain model of canopy reflectance. Agricultural
and Forest Meteorology, 76, 221 – 236.
Kuusk, A. (1995). A fast, invertible canopy reflectance model. Remote Sensing
of Environment, 51, 342 – 350.
Lagacherie, P., Collin-Bellier, C., & Goma-Fortin, N. (2001). Analysing rate
and spatial variability of vinestock mortality in a Languedocian vineyard
from high resolution aerial photographs. Journal International des Sciences
de la Vigne et du Vin, 35(3), 141 – 148.
Lamb, D. W., Weedon, M. M., & Bramley, R. G. V. (2004). Using remote
sensing to predict grape phenolics and colour at harvest in a Cabernet
Sauvignon vineyard: Timing observations against vine phenology and
optimising resolution. Australian Journal of Grape and Wine Research,
10(1), 46 – 54.
Lanjeri, S., Melia, J., & Segarra, D. (2001). A multi-temporal masking
classification method for vineyard monitoring in central Spain. International Journal of Remote Sensing, 16, 3167 – 3186.
le Maire, G., François, C., & Dufrêne, E. (2004). Towards universal broad
leaf chlorophyll indices using PROSPECT simulated database and
hyperspectral reflectance measurements. Remote Sensing of Environment,
89, 1 – 28.
Lichtenhaler, H. K., Lang, M., Sowinska, M., Heisel, F., & Mieh, J. A. (1996).
Detection of vegetation stress via a new high resolution fluorescence
imaging system. Journal of Plant Physiology, 148, 599 – 612.
Li-Cor Inc. (1984), Li-Cor model 1800-12 integrating sphere instruction
manual (Revision 1984). Li-Cor Incorporated, 4421 Superior Street, PO
Box 4425 Lincoln Nebraska, USA, Publication No. 8305-0034.
Marschner, H., Romheld, V., & Kissel, M. (1986). Different strategies in higher
plants in mobilization and uptake of iron. Journal of Plant Nutrition, 9,
695 – 713.
Montero, F. J., Meliá, J., Brasa, A., Segarra, D., Cuesta, A., & Lanjeri, S.
(1999). Assessment of vine development according to available water
resources by using remote sensing in La Mancha, Spain. Agricultural Water
Management, 40, 363 – 375.
O’Neill, N. T., Zagolski, F., Bergeron, M., Royer, A., Miller, J. R., &
Freemantle, J. (1997). Atmospheric correction validation of CASI images
287
acquired over the BOREAS southern study area. Canadian Journal of
Remote Sensing, 23, 143 – 162.
Peñuelas, J., Filella, I., Lloret, P., Muñoz, F., & Vilajeliu, M. (1995).
Reflectance assessment of mite effects on apple trees. International Journal
of Remote Sensing, Vol. 16 – 14, 2727 – 2733.
Peñuelas, J., Gamon, J. A., Fredeen, A. L., Merino, J., & Field, C. B. (1994).
Reflectance indices associated with physiological changes in nitrogenand water-limited sunflower leaves. Remote Sensing of Environment, 48,
135 – 146.
Peñuelas, J., Piñol, J., Ogaya, R., & Filella, I. (1997). Estimation of plant water
concentration by the reflectance water index (R900/R970). International
Journal of Remote Sensing, 18, 2869 – 2875.
Qi, J., Chehbouni, A., Huete, A. R., Keer, Y. H., & Sorooshian, S. (1994). A
modified soil vegetation adjusted index. Remote Sensing of Environment,
48, 119 – 126.
Rondeaux, G., Steven, M., & Baret, F. (1996). Optimization of soil-adjusted
vegetation indices. Remote Sensing of Environment, 55, 95 – 107.
Rougean, J. -L., & Breon, F. M. (1995). Estimating PAR absorbed by
vegetation from bidirectional reflectance measurements. Remote Sensing
of Environment, 51, 375 – 384.
Rouse, J. W., Haas, R. H., Schell, J. A., Deering, D. W., & Harlan, J. C. (1974).
Monitoring the vernal advancements and retrogradation of natural vegetation. NASA/GSFC final report. MD, USA’ Greenbelt 371 pp.
Schultz, H. R. (1996). Leaf absorptance of visible radiation in Vitis vinifera L.:
Estimates of age and shade effects with a simple field method. Scientia
Horticulturae, 66(1-2), 93 – 102.
Sims, D. A., & Gamon, J. A. (2002). Relationships between leaf pigment
content and spectral reflectance across a wide range of species, leaf
structures and developmental stages. Remote Sensing of Environment, 81,
337 – 354.
Tagliavini, M., & Rombolà, A. D. (2001). Iron deficiency and chlorosis in orchard and vineyard ecosystems. European Journal of Agronomy, 15, 71 – 92.
Vogelmann, J. E., Rock, B. N., & Moss, D. M. (1993). Red edge spectral
measurements from sugar maple leaves. International Journal of Remote
Sensing, 14, 1563 – 1575.
Wallace, A. (1991). Rational approaches to control of iron deficiency other than
plant breeding and choice of resistant cultivars. Plant Soil, 130, 281 – 288.
Wellburn, A. R. (1994). The spectral determination of chlorophylls a and b, as
well as total carotenoids using various solvents with spectrophotometers of
different resolutions. Journal of Plant Physiology, 144, 307 – 313.
Zarco-Tejada, P. J., Berjón, A., Morales, A., Miller, J. R., Agüera, J., Cachorro,
V., et al. (2003). Leaf biochemistry estimation on EU high-value crops with
ROSIS and DAIS hyperspectral data and radiative transfer simulation. 3rd
EARSeL workshop on imaging spectroscopy (pp. 597 – 602). Germany’
Munich.
Zarco-Tejada, P. J., Miller, J. R., Mohammed, G. H., & Noland, T. L. (2000).
Chlorophyll fluorescence effects on vegetation apparent reflectance: I. Leaflevel measurements and simulation of reflectance and transmittance spectra.
Remote Sensing of Environment, 74(3), 582 – 595.
Zarco-Tejada, P. J., Miller, J. R., Mohammed, G. H., Noland, T. L., & Sampson,
P. H. (2001). Scaling-up and model inversion methods with narrow-band
optical indices for chlorophyll content estimation in closed forest canopies
with hyperspectral data. IEEE Transactions on Geoscience and Remote
Sensing, 39(7), 1491 – 1507.
Zarco-Tejada, P. J., Miller, J. R., Morales, A., Berjón, A., & Agüera, J. (2004).
Hyperspectral indices and model simulation for chlorophyll estimation in
open-canopy tree crops. Remote Sensing of Environment, 90(4), 463 – 476.
Zarco-Tejada, P. J., Whiting, M., & Ustin, S. L. (2005). Temporal and spatial
relationships between within-field yield variability in cotton and highspatial hyperspectral remote sensing imagery. Agronomy Journal, Vol.
97(No. 3), 641 – 653.

Vineyards - Quantalab

Transcription

Similar documents

Autumn Leaf - Reiman Gardens

Data Sheet - Europa Leisure

Cassia fistula (Golden

Plant - Chaotic gardening

Koelreuteria paniculata (Goldenrain tree, varnish

Understanding the effect of lighting in images

Angiosperm Latin Name: Cyperus involucratus Common Name

Dining room Functionality

available - Cornell University Program of Computer Graphics

metal leaf - Campbell`s Art Class